Enhancing frequent location privacy-preserving strategy based on geo-Indistinguishability

Abstract

The increasing use of hand-held devices which have access to location information, has raised the risk of privacy disclosure. To implement privacy protection on the locations with plenty of check-ins, this thesis proposes a novel location perturbation method based on geo-indistinguishability, which has less quality loss and high privacy guarantee. In order to tackle the problem of how to preserve each person’s frequently occurring position points, we reformulate this issue with a three-step framework. First, the location set is classified by the density-based clustering algorithm, and the privacy budget allocation function is used to allocate the corresponding budget for each cluster. Second, the real location is disturbed according to geo-indistinguishability, and the spanner structure is introduced to increase the efficiency of noise addition and the availability of location data. Finally, we present a privacy metric approach derived from the information entropy to quantify the information leakage by the mechanism, which provides the basis for the analysis of information loss. The experiments are carried out in two real datasets: GeoLife and Taxi GPS reports. Our evaluation confirms that the performance of the proposed strategy is superior to the state-of-the-art solutions in terms of quality loss and privacy metric.